An efficient deep neural network accelerator using controlled ferroelectric domain dynamics

Abstract

Abstract The current work reports an efficient deep neural network (DNN) accelerator, where analog synaptic weight elements are controlled by ferroelectric (FE) domain dynamics. An integrated device-to-algorithm framework for benchmarking novel synaptic devices is used. In poly(vinylidene fluoride-trifluoroethylene)-based ferroelectric tunnel junctions (FTJs), analog conductance states are measured using a custom pulsing protocol, and associated control circuits and array architectures for DNN training are simulated. Our results show that precise control of polarization switching dynamics in multi-domain polycrystalline FE thin films can produce considerable weight-update linearity in metal–ferroelectric–semiconductor (MFS) tunnel junctions. Ultrafast switching and low junction currents in these devices offer extremely energy-efficient operation. Via an integrated platform of hardware development, characterization and modeling, we predict the available conductance range, where linearity is expected under identical potentiating and depressing pulses for efficient DNN training and inference tasks. As an example, an analog crossbar-based DNN accelerator with MFS junctions as synaptic weight elements showed >93% training accuracy on a large MNIST handwritten digit dataset while, for cropped images, >95% accuracy is achieved. One observed challenge is the rather limited dynamic conductance range while operating under identical potentiating and depressing pulses below 1 V. Investigation is underway to improve the FTJ dynamic conductance range, maintaining the weight-update linearity under an identical pulse scheme.